Mature leaves of Arabidopsis contain numerous stomata interspersed with non-stomatal cells (1
). Stomata derive from precursors (meristemoids) that may undergo several rounds of asymmetric division during which only one daughter retains meristemoid identity, suggesting that they behave as stem cells for a limited period (2
). Meristemoids eventually differentiate into Guard Mother Cells (GMCs) which in turn produce the two guard cells of the stomatal pore. The molecular genetic and cellular control of many of these steps has been extensively analysed (4
). However, it is still unclear how initial stomatal lineage patterns are generated. This is because previous studies have only imaged gene expression at particular time points, or followed lineages after many stomata have already formed.
To understand how stomatal patterning occurs, we used live confocal imaging methods to track growing leaves expressing various markers from precursor initiation to differentiation. We first tracked a plasma membrane GFP marker line to determine early lineage patterns. Retrospective analysis of the tracking data was used to classify cells that went on to produce GMCs. A cell that included a single GMC descendant during the period of imaging was termed a precursor cell or P cell, while all other cells were called N cells. We tracked 50 lineages distributed across the leaf blade. The results showed that P cells behave as stem cells for 3–4 rounds of division, after which they become GMCs ( and S7
Fig. 1 Fate and specification of stomatal precursors. (A) A Tracked lineage showing P cells (red), GMC cells (A, blue), guard cells (G, blue), and N cells (yellow). Times of image capture (to the nearest hour) from left to right are: 00h, 04h, 20h, 33h, 47h, (more ...)
P cells exhibit several properties that distinguish them from N cells. P cells divide more often than N cells and become progressively smaller through division. The reduction in size is partly because P cells divide asymmetrically, with the smaller cell usually retaining P cell identity and being 39% of the parent cell size on average (Fig. S1
). Reduction in P cell size is also enhanced by P cells tending to divide at successively smaller sizes (). N cells may also divide, but such divisions are usually symmetric and occur when N cells reach more than twice their birth size (Fig. S2
). In some cases N cells produce daughters that divide asymmetrically at less than twice their birth size, suggesting these N cells have reacquired P identity (, S3
As well as showing limited stem cell behaviour, stomatal lineages also exhibit stereotypical patterns of division ( and S7
). The most common pattern is for P cells to undergo 2–3 rounds of division in alternating orientations, followed by a division across adjacent walls, creating a spiral arrangement with an internalised P cell (, and S7A, B
). The second pattern is similar except that two successive divisions span adjacent walls ( and S7C
). The third and least common pattern involves a parallel rather than alternating division ( and S7D
). Lineages where the P cell was not internalised were largely truncated versions of these arrangements (Fig. S8
). Any model for stomatal patterning therefore needs to account for both the stem cell behavior of P cells and the stereotypical manner in which they divide.
Fig. 2 Patterns observed among lineages involving more than two rounds of division (47 lineages analyzed). Interpretive diagrams (A,C,E) are shown above time-lapse confocal images (B,D,F). (A) P cell undergoes 2–3 divisions in alternating orientations, (more ...)
To account for stem cell property of P cells, we hypothesised that they express an identity factor that is only maintained in a single daughter following division, allowing it to retain P cell identity. A good candidate gene for this factor is SPEECHLESS (SPCH). SPCH encodes a bHLH transcription factor that is required for the formation of stomatal lineages and is also able to drive stomatal development when activated ectopically (4
). SPCH protein is found in a subset of epidermal cells, often adjacent to each other, and is restricted to smaller cells in older leaves (4
). Although maintenance of SPCH in single daughters is consistent with these observations, this conclusion was not previously drawn, perhaps due difficulty in inferring such a behavior from static images. We therefore tracked leaves containing a functional SPCH::SPCH-GFP protein fusion for extended periods to determine its spatiotemporal dynamics. SPCH protein was detected in P cells and in both daughters immediately following P division, accounting for the observation that SPCH is often found in adjacent cells ( and S4
). However, a few hours after division, SPCH protein was no longer seen in the N cell daughter, while it was maintained in the P daughter ( and S4
). This pattern of maintenance in single daughters continued through successive rounds of division until formation of the GMC, when SPCH expression was extinguished. These results therefore support the hypothesis that P identity is maintained in single daughters, and show that this identity is most likely mediated by SPCH activity.
This model predicts that SPCH should be reactivated in N cells that regain P identity. Live imaging of the SPCH-GFP line showed that some N cells regain SPCH protein expression at later stages. After regaining SPCH signal, these cells behave as P cells - dividing asymmetrically with SPCH activity being maintained in a single daughter cell ( and S5A, C
). Thus, cells that regain SPCH activity also regain P identity as predicted. As well as being reactivated in these early lineages, SPCH could also be reactivated at later stages of development in cells neighbouring stomata, giving rise to secondary stomata (3
) (Fig. S5E, G
). A further prediction of the model is that cells in a spch
mutant leaf should behave as N cells. Consistent with this prediction, tracking spch
leaves revealed bifurcating rather than stem cell lineages, and daughter cells that divided when they reached more than twice their birth size, typical of N cells ( and S6
Having established that P cell identity is likely maintained in single daughters through SPCH, we extended our model to investigate how P cells divide asymmetrically and become surrounded by their N cell relatives. A key gene involved in controlling asymmetric division of stomatal precursors is BASL (11
). BASL protein is localised to the nucleus and to a peripheral region of the cell; the peripheral pool is highly polarized, is sufficient for BASL function and is inherited by the larger daughter cell following division (11
). The smaller daughter may regain peripheral BASL expression, which again marks a region that will belong to the larger daughter following the next division. Although these results show that peripheral BASL acts as a polarity marker for an individual asymmetric division, it is unclear how BASL’s peripheral location and/or polarizing activity operates during consecutive divisions, or how this process is related to the observed stereotypical division patterns.
To incorporate BASL in our model, we first considered a simplified case of a growing one-dimensional file generated by dividing P cells. Two basic patterns can arise depending on how BASL is positioned over multiple division rounds.
In one case, peripheral BASL is localised near the new division boundary following division. This rule leads to a lineage with an apical P cell (), and is formally equivalent to a variety of plant and animal stem cells, in which stem cell identity is maintained at one end (12
Fig. 3 Models for P cell lineages. (A–B) One dimensional lineage models. (A) Following division, the cell without peripheral BASL (green) keeps P identity (red) while the other cell becomes an N cell. Peripheral BASL in the new P cell becomes localized (more ...)
Another possibility is that the peripheral BASL is localised away from the new division boundary () (14
). In this case polarity of the daughter is opposite to that of the mother, so we call this pattern post-mitotic polarity switching
. Positioning of BASL may be determined either through a signal from the new boundary as in yeast (15
), or an inhibitory signal from the recent non-stem cell neighbour (, orange). Successive application of the post-mitotic polarity switching rule leads to internalisation of the P cell, consistent with the observed stomatal lineages.
To determine whether polarity switching could account for the observed stereotypical stomatal lineage patterns, we extended our model to two dimensions. This extension required a rule for determining the position and orientation of new division walls. In accordance with previous studies, cells were divided by walls taking the shortest path across the cell passing through the nucleus, initially assumed to be at the cell centre (17
) (). Asymmetric divisions could then be implemented by displacing the nucleus away from the centre of the cell. We show that small displacements result in geometrically asymmetric divisions (), while larger displacements produce triangular cells (), a feature exhibited by P cells during later division rounds ( and S7
). A post-mitotic polarity switch was then incorporated by hypothesising that BASL protein becomes localised in the cell periphery at a position furthest removed from the newest wall following P cell division. This model did not consistently lead to internalisation of P cells (). To address this problem, we explored a second model in which peripheral BASL becomes localised at a position furthest removed from all recent division walls (19
) (i.e., all division walls formed after the establishment of P cell identity in the initial cell). This model generated internalised P cells ().
To determine whether the second model could account for all observed lineage patterns (), we varied the initial cell geometry, anisotropy of growth and extent to which the nucleus is displaced away from peripheral BASL. This parameter search showed that all experimentally observed patterns could be generated by the model (, and ).
Fig. 4 Model validation. (A,B) By altering the initial cell geometry, growth rates and/or the amount that peripheral BASL displaces the nucleus, lineage patterns are obtained that correspond to those in . (C, F, I) Location of peripheral BASL (more ...)
As a further test of the model we tracked the position of BASL in P cell lineages over multiple rounds of division (, S9 and S10
). In accordance with previous studies (11
), peripheral BASL always marked the region that will belong to the N cell in the next division. From the cell outlines and sequence of divisions we computed the position furthest from recent division walls, corresponding to the position of peripheral BASL predicted by the model. In 13 out of 21 divisions analysed, peripheral BASL location was correctly predicted (labelled (i) in and S9
). In 5 cases the predicted and observed BASL positions were slightly discordant (divisions labelled (ii) in and S9
). Such discrepancies are not surprising given the very simplified assumptions used for making the predictions, such as the new walls acting perfectly uniformly to repel BASL. In the remaining 3 cases, peripheral BASL was located at the opposite end to that predicted (marked (iii) in and S9
). These exceptions arise when the daughter P cell has near bilateral symmetry about a plane perpendicular to the new division wall. For a perfectly symmetrical condition, there can be two locations that are furthest from the new wall, on opposite sides the cell, so either of these could acquire peripheral BASL. Thus, small stochastic fluctuations may account for peripheral BASL sometimes being localised at the opposite end to that predicted ().
As a final validation of the model, we used it to simulate entire stomatal lineages using the initial P cell’s geometry (simplified with straight walls), growth parameters and initial location of peripheral BASL. In many cases, the model could account for both the observed pattern of divisions and location of peripheral BASL over multiple rounds of division ( and Fig. S10
). In symmetric cases where peripheral BASL is predicted to be at the opposite corner to that observed, the model predicts the lineage to spiral opposite to the observed direction, as expected ( and Fig S10
Post-mitotic polarity switching provides a simple mechanism by which plant stem cells may generate their own neighbourhood while also spacing themselves apart. While this mechanism does not address all elements of stomatal spacing, such as activation of SPCH during secondary stomata formation, it provides one way of preventing formation of adjacent stomata. Polarity switching might be extended in several ways. The stem cell could maintain its character for longer than observed for P cells, generating greater levels of internalisation, as observed in the MUTE mutant in which GMCs fail to form (10
). Polarity switching over longer periods might also underlie the behaviour of fern and moss apical cells which continually generate neighbours around them while maintaining their stem cell identity (20
Various elements of the polarity switching mechanism are also found for animal stem cells. Polarity of division plays a key role in maintaining animal stem cell identity (23
). Animal stem cells may also generated their own neighbourhoods (niches), as described for the Drosophila
) and mouse intestine (25
). However, unlike stomatal development, cell rearrangements play an important part, so it is unclear whether polarity switching would provide a robust mechanism for patterning in these cases. Thus, while sharing many elements with animal systems, the fixed nature of plant cells may lead to distinctions in the way cells pattern themselves in dynamically changing tissues.